Communication pubs.acs.org/crystal
Coformer Replacement as an Indicator for Thermodynamic Instability of Cocrystals: Competitive Transformation of Caffeine:Dicarboxylic Acid MHD. Bashir Alsirawan,† Venu R. Vangala,† John Kendrick,‡ Frank J. J. Leusen,‡ and Anant Paradkar*,† †
Centre for Pharmaceutical Engineering Science and ‡School of Chemistry and Forensic Science, University of Bradford, Bradford, West Yorkshire BD7 1DP, U.K. S Supporting Information *
ABSTRACT: The thermodynamic stability of caffeine (CA) cocrystals with dicarboxylic acids (DAs) as coformers was investigated in the presence of a range of structurally related dicarboxylic acids (SRDs). Two experimental conditions (slurry and dry-grinding) were studied for mixing the cocrystal and the SRD additive. The additives oxalic, malonic, and glutaric acid led to the replacement of the acid coformer for certain cocrystals. Interestingly, a change in stoichiometry was observed for the CA:maleic acid system. A stability order among the cocrystals was established depending on their tendency to replace the coformer. To understand the factors controlling the relative stabilities, lattice energies were calculated using dispersion corrected density functional theory (DFT). Gibbs free energy changes were calculated from experimental solubilities. The observed stability order corroborated well with lattice energy and Gibbs free energy computations.
C
calculating lattice energies for the isomers using an atom− atom potential method.19 Here we report the effect of structurally related dicarboxylic acid (SRD) additives on the stability of cocrystals. The stabilities of caffeine (CA) and dicarboxylic acid (DA) cocrystals are determined in the presence of SRDs, which are hypothesized to compete with the DA coformers and could potentially impact the integrity of the cocrystals. The outcomes of this study provide important input to formulators as the presence of structurally related components is routinely seen in complex and multicomponent formulations. To understand such competitive transformations further, we utilized and compared two methods for predicting whether a system is stable or not. The first method calculates lattice energies after optimizing known crystal structures determined from single crystal X-ray diffraction, using density functional theory.20,21 The second method calculates cocrystal stability from the Gibbs free energy change determined from experimental solubilities.22,23 The selected cocrystals consist of CA with the DAs, oxalic acid (OX), malonic acid (MO), glutaric acid (GL), and maleic acid (ML) (Scheme 1). Cocrystals CA:OX 2:1 (A), CA:MO 2:1 (B), CA:GL 1:1 FII (C), CA:GL 1:1 FI (D), CA:ML 1:1
ocrystal systems are increasingly being considered for many industrial applications including explosives,1 semiconductors,2 pigments,3 and pharmaceuticals.4 Recently, the FDA approved the valsartan:sacubitril cocrystal Entresto as a dosage form in the treatment of heart failure.4 In the past decade there have been reports about the development of innovative technologies for the manufacture and formulation of cocrystals.5,6 To obtain a successful pharmaceutical cocrystal product, it is crucial to understand cocrystal behavior during manufacturing, storage, and administration as cocrystals can undergo different types of structural changes. Recent reports suggested that solvent mediated transformations such as dissociation,7 polymorphism,8 and solvation9 are mainly due to a large solubility7,10 or a small pKa11,12 difference between the molecules involved in cocrystallization. Notably, solventfree thermal or shear force mediated13,14 transformations that include stoichiometric conversion15 have been reported, which were ascribed to the energy differences between the cocrystal and the individual components.15,16 The stability issue in formulations is further complicated due to the presence of additives. Caira et al. have reported selective formation of sulfonamide cocrystals with some aromatic carboxylic acids and demonstrated the role of hydrogen bonding in such preferential cocrystal formation.17 Fischer et al. studied a set of cocrystal competitive grinding reactions and inferred the role of kinetic factors.18 Similarly, the affinity of aminobenzonitrile isomers with a 1,1,6,6-tetraphenylhexa-2,4-diyne-1,6-diol host molecule was also assessed experimentally and mathematically by © XXXX American Chemical Society
Received: March 23, 2016 Revised: May 3, 2016
A
DOI: 10.1021/acs.cgd.6b00458 Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Communication
Scheme 1. Molecular Structures of Caffeine and Coformers
(E), and CA:ML 2:1 (F) were prepared by solution crystallization (see also Supporting Information, SI). The intermolecular hydrogen bonding motif between CA and DAs is the same in all cocrystals.11 CA:DA−SRD binary mixtures were prepared with sufficient SRD to form a cocrystal. Mixtures were split into two lots: the first was transferred to slurries (for aqueous processing) and the second was used for dry grinding (mechanical processing). The aim was to mimic typical formulation processes such as wet granulation, drying, milling, or storage. CA:DAs without SRDs were subjected to aqueous processing and mechanical processing to analyze the effect of processing alone. Structural analysis was performed by powder X-ray diffraction (PXRD). Thermogravimetric analysis (TGA) was used to determine the solvent content of all binary mixtures before subjecting them to mechanical processing (see SI). The results are summarized in Table 1. Slurry processing revealed that most of the cocrystals undergo coformer replacement in the presence of SRDs. In addition, the presence of ML with F caused a change in stoichiometry and the formation of E. For cocrystals processed without SRD (NA samples), no change in cocrystal composition was observed except for F and D where dissociation and polymorph transformation into C occurred, respectively. The stability order, based on the ability of the cocrystal to withstand replacement of its original coformer or any other transformation, was A > B > C > E > F ≈ D. The stability order obtained by dry grinding was A > B > C ≈ E > F > D, which is similar to the order obtained with slurry processing. Figures 1 and 2 show the PXRD patterns of C and F after aqueous processing in the presence of SRD. The PXRD patterns of the other cocrystals after aqueous processing and mechanical processing are provided in the SI. To better understand the process of coformer replacement, consider the following equilibrium: XaYb + c Z ⇌ XaZc + b Y
Figure 1. PXRD patterns for C after aqueous processing in the presence of SRDs (1) exptl C, (2) C slurry, (3) C + ML slurry, (4) C + GL slurry, (5) simulated (sim) C + ML slurry, (6) C + MO slurry, (7) sim B, (8) C + OX slurry, and (9) sim A.
where XaYb is the starting cocrystal, Z is the additive, XaZc is the cocrystal after coformer replacement, Y is the coformer replaced by the additive, and a, b, and c are numbers of molecules. A negative value of ΔG for the equilibrium indicates that the cocrystal containing the new coformer is thermodynamically more stable than the original cocrystal, i.e., the reaction favors coformer replacement. The current study has explored two methods to calculate energy changes for replacement reactions. The first involves DFT calculations, and the second uses experimental equilibrium solubility data. The lattice energies of the cocrystals computed by a dispersion corrected DFT method (DFT-D, see SI) were used to estimate the relative stabilities of the cocrystals in the presence of SRDs using eq 2, where E(X) refers to the lattice energy of X: ΔE = E(XaZc) + bE(Y) − [E(XaYb) + cE(Z)]
(1)
(2)
Table 1. Results of Mixing SRD with CA:DA Cocrystals Using Slurry and Dry Grinding (DG) SRD
CA:DA
slurry
DG
CA:DA
slurry
DG
CA:DA
slurry
DG
OX MO GL ML NAa OX MO GL ML NAa
A
A A A A A A B C C C
A A A A A A B C C C
B
A B B B B A B C E E
A B B B B A B E E E
C
A B C C C A B C E CA + MLb
A B C C C A B C E F
D
E
F
a
Cocrystals processed without SRD. bCocrystal dissociated into its native components. For ready reference, cocrystals A = CA:OX 2:1, B = CA:MO 2:1, C = CA:GL 1:1 FII, D = CA:GL 1:1 FI, E = CA:ML 1:1, and F = CA:ML 2:1. B
DOI: 10.1021/acs.cgd.6b00458 Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Communication
Table 3. CA:DA Solubilities, Ksp, and pH Values of Saturated Solutions in Water cocrystal
code
CA Sa
DA Sa
Kspb
pH
CA:OX CA:MO CA:GL FII CA:GL FI CA:ML 1:1 CA:ML 2:1
A B C D E F
0.06 0.16 0.31 0.47 0.30 0.25
0.02 0.38 0.78 1.070.38 0.49
8.3 × 10−05 9.2 × 10−3 0.24
1.8 1.5 2.3
0.15
1.5
a
Unit is M. bUnit is M2 for 1:1 cocrystals and M3 for 2:1 cocrystals. Blank spaces show results that could not be obtained due to experimental difficulties.
Table 4. Energy Changes (kcal/mol) for Coformer Replacement Reactions and Cocrystal Integrity Observations Whether Replacement Is Taking Place (√) or Not (X) under Slurry Processing starting cocrystal
The second method for calculating the free energy change in the equilibrium (reaction 1) uses eq 3, which follows previous work.22,23 For details of derivations see the SI.
a
(3)
where Ksp is the solubility product for the cocrystal and S is the equilibrium solubility of the pure components. Four general replacement reactions are proposed Type 2: X 2Y + Z ⇌ X 2Z + Y
Type 3: 2XY + Z ⇌ X 2Z + 2Y Type 4: X 2Y + Z ⇌ XZ + Y + X
Ksp and S in water for CA cocrystals and their pure components were determined (Tables 2 and 3). The shaking flask method followed by HPLC analysis was used for measurements (see SI). Results for A and E are presented in Table 4. The data for the rest of the cocrystals are listed in the SI.
S (M)
pH
pKa11
CA OX MO GL ML
20.32 116.7 741.5 531.4 389.1
0.10 1.30 7.13 4.02 3.35
2.6 1.2 0.5 1.3 1.1
3.6 1.3 2.8 4.3 1.8
ML
E
ΔE
ΔGEx °
replacement
−5.79 −6.62 −8.44 −6.91 −9.18 8.11 2.32 0.15 −0.76 −1.07
−7.50 −10.06
√ √ √ √ √ X X X X √
−8.53 10.68 3.18 −0.76
Type 1. bType 2. cType 3. dType 4.
CA:DA
slurry
ΔERe
ΔGRe °
ΔEFo
ΔS (M)
ΔpKa
A B C D E F
1 2 3 5 4 5
1 2 3 5 4 6
1 2 4
−9.3 −3.5 −1.3 −0.4 −1.2 −0.1
1.20 7.03 3.92 3.92 3.25 3.92
2.3 0.8 0.7 0.7 1.8 0.7
3
In the same way, we found that CA:ML has two stoichiometries with distinct stabilities. Therefore, pKa or solubility differences cannot explain the differences among polymorphs or stoichiometries. Instead, thermodynamic parameters such as Gibbs free energies and lattice energies show a robust relationship with experimental findings (Table 5). The stability order obtained by the DFT-D calculations is A > B > C > E > D > F, which follows the same trend as the stability order obtained by experimental methods. The stability order obtained by ΔGRe ° calculations is A > B > E > C > F > D.
Table 2. Aqueous Solubilities, pH, and pKa Values at 25 °C of Saturated Solutions of Pure CA and DAs S (mg/mL)
A
Table 5. CA:DAs Stability Orders for Slurry, DFT Method (ΔERe), Solubility Data Method (ΔG°Re), and Stability Relative to Original Components (ΔEFo) Compared with S and pKa Difference, ΔS and ΔpKa, Respectively
Type 1: XY + Z ⇌ XZ + Y
material
OX
There is no correlation between pKa or solubility difference and cocrystal tendency to destabilize by coformer replacement (Table 5). Both were investigated by Trask and Jones11 as the CA:GL system exhibits two polymorphs having profoundly different durabilities at high %RH levels.
K spXYSZc K spXZS Yb
result cocrystal
Bb Cc Dc Ec Fb Ad Bd Ca Da Fd
Figure 2. PXRD patterns for F after aqueous processing in the presence of SRDs (1) exptl F, (2) sim F, (3) F slurry, (4) CA + ML 2:1 physical mixture, (5) F + ML slurry, (6) sim E, (7) F + GL slurry, (8) sim C, (9) F + MO slurry, (10) sim B, (11) F + OX slurry, and (12) sim A.
° = −RT ln ΔG Re
SRD
C
DOI: 10.1021/acs.cgd.6b00458 Cryst. Growth Des. XXXX, XXX, XXX−XXX
Crystal Growth & Design
Communication
Notes
The differences between the slurry and dry grinding results indicate that mechanical processing delivers lower energy than aqueous processing. Therefore, dry grinding may not reach the kinetic barrier for some samples. This might explain why the replacement did not occur for E in the presence of GL during mechanical processing, whereas the coformer replacement did take place during aqueous processing. Coformer replacement was easiest in D and F. However, the stability order between the two systems is unclear as D converts to the more stable polymorph C if slurried in the presence of ML instead of undergoing coformer replacement. This is supported by the lattice energy calculation results. The lattice energy change of D relative to its components (−0.4 kcal/mol, Table 2) is larger than for F (−0.1 kcal/mol). However, C is more stable (−1.3 kcal/mol). Moreover, the ΔE value of polymorph conversion from D to C (−0.91 kcal/mol, Table S6) is larger than the stability of D. ΔG° values calculated using S data resulted in a stability order that is similar to both the experimental and the computational order, except that E and F are slightly more stable than C and D, respectively. This might be due to the fact that computational stability values (Table 5) of these systems are very similar; −1.3 and −1.2 kcal/mol for C and E, respectively, and −0.4 and −0.1 kcal/mol for D and F, respectively. Moreover, S values for D and F could not be reliably measured; therefore, S and ΔG°Re data is not mentioned (Tables 3 and 4). In conclusion, the presence of a structurally similar additive during cocrystal destabilization, whether it is solvent or solventfree mediated, may have a profound effect on stability. Therefore, it is advisable to perform stability testing for the mixtures of cocrystal and additives to be formulated in a product. Lattice energy or Gibbs free energy calculations employing computational or solubility methods can predict the propensity of a cocrystal to destabilize, including coformer replacement. Coformer replacement of a given cocrystal can be deliberately stimulated to test the coherence of the cocrystal even for coformers similar in binding strength. Both aqueous processing and mechanical processing involve components reaching a higher energy level. Therefore, kinetic experiments, such as quantitative phase analysis, can be performed to obtain a deeper knowledge about the mechanism.
■
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS Authors would like to thank EPSRC (EP/J003360/1, EP/ L027011/1) and UKIERI (TPR/23). V.R.V. thank RSC researcher mobility grant. Authors are thankful to Prof. Ashwini Nangia and Dr. C. Malla Reddy for useful discussion about cocrystal stability.
■ ■
ABBREVIATIONS HPLC, high performance liquid chromatography; DFT, density functional theory; S, solubility; SC, solution crystallization
ASSOCIATED CONTENT
S Supporting Information *
The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.cgd.6b00458. Cocrystal preparation, characterization, experimental section, density functional theory method and further details, solubility determination, and Gibbs free energy calculations (PDF)
■
REFERENCES
(1) Landenberger, K. B.; Bolton, O.; Matzger, A. J. Angew. Chem., Int. Ed. 2013, 52, 6468−6471. (2) Sokolov, A. N.; Friscić, T.; MacGillivray, L. R. J. Am. Chem. Soc. 2006, 128, 2806−2807. (3) Bučar, D.-K.; Filip, S.; Arhangelskis, M.; Lloyd, G. O.; Jones, W. CrystEngComm 2013, 15, 6289−6291. (4) FDA Approves New Drug to Treat Heart Failure. http://www. fda.gov/NewsEvents/Newsroom/PressAnnouncements/ucm453845. htm. (5) Schultheiss, N.; Newman, A. Cryst. Growth Des. 2009, 9, 2950− 2967. (6) Qiao, N.; Li, M.; Schlindwein, W.; Malek, N.; Davies, A.; Trappitt, G. Int. J. Pharm. 2011, 419, 1−11. (7) Eddleston, M. D.; Thakuria, R.; Aldous, B. J.; Jones, W. J. Pharm. Sci. 2014, 103, 2859−2864. (8) Trask, A. V.; Motherwell, W. D. S.; Jones, W. Chem. Commun. 2004, 7, 890−891. (9) Jayasankar, A.; Roy, L.; Rodríguez-Hornedo, N. J. Pharm. Sci. 2010, 99, 3977−3985. (10) Childs, S. L.; Rodriguez-Hornedo, N.; Reddy, L. S.; Jayasankar, A.; Maheshwari, C.; McCausland, L.; Shipplett, R.; Stahly, B. C. CrystEngComm 2008, 10, 856−864. (11) Trask, A. V.; Motherwell, W. D. S.; Jones, W. Cryst. Growth Des. 2005, 5, 1013−1021. (12) Trask, A. V.; Motherwell, W. D. S.; Jones, W. Int. J. Pharm. 2006, 320, 114−123. (13) Eddleston, M. D.; Lloyd, G. O.; Jones, W. Chem. Commun. 2012, 48, 8075−8077. (14) Goud, N. R.; Nangia, A. CrystEngComm 2013, 15, 7456−7461. (15) Karki, S.; Friscic, T.; Jones, W. CrystEngComm 2009, 11, 470− 481. (16) Arhangelskis, M.; Lloyd, G. O.; Jones, W. CrystEngComm 2012, 14, 5203−5208. (17) Caira, M. R.; Nassimbeni, L. R.; Wildervanck, A. F. J. Chem. Soc., Perkin Trans. 2 1995, 12, 2213−2216. (18) Fischer, F.; Joester, M.; Rademann, K.; Emmerling, F. Chem. Eur. J. 2015, 21, 14969−14974. (19) Caira, M. R.; Nassimbeni, L. R.; Toda, F.; Vujovic, D. J. Am. Chem. Soc. 2000, 122, 9367−9372. (20) Neumann, M. A.; Leusen, F. J. J.; Kendrick, J. Angew. Chem., Int. Ed. 2008, 47, 2427−2430. (21) Chan, H. C. S.; Kendrick, J.; Leusen, F. J. J. Angew. Chem., Int. Ed. 2011, 50, 2979−2981. (22) Schartman, R. R. Int. J. Pharm. 2009, 365, 77−80. (23) Maheshwari, C.; Jayasankar, A.; Khan, N. A.; Amidon, G. E.; Rodríguez-Hornedo, N. CrystEngComm 2009, 11, 493−500.
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Author Contributions
M.B.A., V.R.V., and A.P. have contributed with the idea, design, and performing the experimental work that includes slurry and grinding processes, Gibbs free energy calculations, and write up of the manuscript. J.K. and F.J.J.L. have performed DFT computations and contributed to the manuscript preparation. D
DOI: 10.1021/acs.cgd.6b00458 Cryst. Growth Des. XXXX, XXX, XXX−XXX
